[split] Verbose cipher?

[MarcoP]

I used a simple simulated-annealing Python script to experiment with verbose ciphers with English as plain-text language. I omitted the rare plain-text characters Q,X,Z. As already discussed, with this approach one has to take liberties with spaces.
This is one of the results for the first lines of f76r:

A -> ['op', 'che', 'oa']
B -> ['cth']
C -> ['in']
D -> ['sh', 'ai']
E -> ['q', 'k', 'r']
F -> ['chk']
G -> ['as']
H -> ['y', 'eo']
I -> ['ain']
J -> ['ee']
K -> ['cph']
L -> ['l', 'or']
M -> ['ol']
N -> ['cho', 'of']
O -> ['chcfh', 'ed', 'chcph']
P -> ['ea']
R -> ['i', 'al']
S -> ['ey', 's']
T -> ['d', 'p', 'chd']
U -> ['ot']
V -> ['am']
W -> ['ok']
Y -> ['ar']

The deciphering of the first line results in a list of actual English word, but things already break down significantly at line two.

Code:

tune          lot   had   them  to    hot    had    send  hath    sigh TUNEL        OTH    ADTH  EMT    OH    OTH    ADS    ENDH  ATH    SI  GH potchokor chcfhdy opshdy qolp chcphy chcphdy opshey qofshy opchdy sain asy       do              by    at                      do  hem TDOH  EUTTHY    BY    AT  TI  WI    EWHE    DOH  EM  I    DPLSEWJH dshedy qotddyar cthar chep dain okain qokeor shedy qol ain shealsqokeey

     

This is the cipher of a short English sentence:

Code:

  I  N TH E  ain chody r B    E G I  N N I  N G G       O D C  RE A  TE  DT H EHE  A V EN cth qasain ofchoain ofasas chcphshin alkopchdr aipeo qykcheam rcho  A N DT H EE  A RTH oaofshdeo qr opaldy

Some thoughts:

• In the absence of grammar, a word salad of a few consecutive words is not significant. Also, a deciphering where the spelling of all longer words requires some tweaking is highly suspicious; in other words, an alternative to searching for perfect matching short words is searching for loosely matching longer words. For instance, the sequence DPLSE at the end of the second line could stand for "DISPLEASE": this is not convincing, e.g. why was the first S dropped and the second one preserved?
  
• In the absence of grammar, short words are not significant. The probability of randomly getting a longer word is exponentially lower than that of getting short words. Again, longer words matches are only significant if they follow clear rules with no tweaking: these rules may not correspond to modern normalized spelling, e.g. the Latin word "consummatus" may appear as "9sumat9" in a medieval manuscript, the rules mapping the two forms into each other are clear (positional abbreviation symbols and dropping of double consonants).
  
• Grammar makes a sequence of short words significant: "and if ye go to war in your land" is not the same as "in ye war your land and to go if".
  
• A verbose cipher does not easily reproduce Voynichese word structure. Probably something like You are not allowed to view links. Register or Login to view. would be needed: each plain-text character needs at least three different encodings corresponding to the three layers of Voynichese words prefix-core-suffix.